Sam Altman says one of AI’s underappreciated risks is that people may let the technology do so much thinking for them that their own mental muscles begin to weaken.
“One that I don’t think gets much attention is how are we going to avoid cognitive atrophy?” the OpenAI CEO said on an episode of the “Invest Like the Best” podcast that aired on Tuesday, when asked about the most important unresolved questions facing AI.
“How are we going to use these tools and make sure that we are stretching our brains more and more and continuing to understand the stuff that really matters?” Altman said.
His comments come as a growing number of researchers have warned that while the technology could make workers faster, it could also weaken their underlying job and cognitive skills.
“There’s never been a tool like this that has offered us the opportunity to let thinking be done for us,” Nataliya Kosmyna, an MIT researcher whose preliminary study found that people using generative AI to write essays performed worse over time than participants using Google or no aid, told Business Insider earlier this month.
Vivienne Ming, a theoretical neuroscientist and AI researcher, has described the risk of people abandoning cognitive work because AI can do it for them as “chronic substitution.” She told Business Insider last month that repeated substitution could weaken cognitive reserve, a key defense against dementia.
Those concerns are especially pressing as AI tools move beyond answering questions and generating text to taking on increasingly complex work, including document review, research, analysis, customer support, and coding.
Altman isn’t the only tech leader to have raised concerns about AI’s effect on cognition.
Last year, Arthur Mensch, the CEO of Paris-based Mistral AI, told The Times that AI’s biggest risk may be “deskilling” workers and making people progressively lazier as they increasingly rely on the technology to search for information.
Still, during the podcast interview, Altman struck a more positive tone about what lies ahead.
He described a “pretty magical” future where a personal AI could be always on, reading documents, listening to meetings, recalling emails from weeks earlier, and using spare computing power overnight to generate ideas and suggest what they should do next.
However, he said that even powerful models may struggle to replicate the qualities that make humans distinctive.
“The world may need like a new kind of word for the kind of judgment that people are very good at, that AI seems to really deeply struggle with,” he said.